It’s one of the first questions owners ask us, and it should be. Your clients trust you with their details. That trust extends to anyone, and anything, you let near them.
There isn’t a single yes or no. Safety depends on how a particular system is built: what it can see, where the information goes and who can get to it. You don’t need to be technical to check. You need to ask the right questions and expect clear answers.
This article is general information, not legal advice. For advice on your own obligations, speak to a qualified adviser.
1. What data does it actually need?
Start here, because the safest data is data a system never touches. An automation that drafts invoices needs job details and prices. It probably doesn’t need your clients’ full contact history, their notes or their bank details.
Ask the provider to list the data each automation will read and write. If the answer is ‘everything, just in case’, push back. A well-built automation only connects to the systems and data it needs.
- Which systems will it connect to?
- Which records and fields will it read?
- Will it create, change or delete anything?
2. Where is it stored and processed?
When AI works on your information, the data may pass through more than one service: your own systems, the automation platform and the company that runs the AI model. Each one is a place your data travels.
- Where does each step happen, including which countries?
- Does any service keep a copy, and for how long?
- Is our data used to train AI models?
3. Who can access it?
Think about people as well as systems. Who at the provider can see your data, and why? Who on your own team can see what the automation sees?
Access should match the job. An automation that handles accounts doesn’t need to be visible to everyone in the business, and a provider shouldn’t keep access to your systems once the work no longer needs it.
- Whose login or credentials will it use?
- Can access be limited to specific records or folders?
4. Is there a record of what it did?
If something goes wrong, you need to see what happened. A good system keeps a log: what it read, what it changed, what it sent and when. A log turns ‘we think it worked’ into something you can check.
- Is there a log of every action?
- Who can see it, and how long is it kept?
5. Where does a person check the work?
No automation should be trusted blindly with client information. Ask what it does when something looks unusual: a record that doesn’t match, a request outside the rules, a message from someone it doesn’t recognise.
The answer you want is that it stops and flags it for a person. It shouldn’t guess, and it shouldn’t send client information anywhere new without someone approving it.
6. How do we remove access?
Projects end. Staff change. Providers change. Before you connect anything, know how to disconnect it.
- How do we revoke access to each system?
- What happens to data the provider holds when we stop?
- Can we confirm it has been deleted?
7. How do you handle personal information under the Australian Privacy Principles?
The Australian Privacy Principles, set out in the Privacy Act 1988, cover how personal information is collected, used, disclosed and kept secure, including when it’s sent overseas. Whether and how they apply to your business depends on your circumstances, so it’s worth checking.
Either way, they’re a sensible standard to hold a provider to. Ask how they handle personal information in line with the Australian Privacy Principles, and whether they have a privacy policy you can read. A provider who takes this seriously can explain it plainly.
The checklist
Take these questions into any conversation about connecting AI to client data.
Before you give AI access to client data
- What data will it read, write or change, and is all of it needed?
- Which systems will it connect to, and with whose credentials?
- Where is the data stored and processed, including any overseas services?
- Does any service keep a copy, or use our data to train AI models?
- Who at the provider, and on our team, can see the data?
- Is there a log of what the automation did?
- Where does a person check the work, and what gets flagged?
- How do we remove access, and what happens to our data when we do?
- How does the provider handle personal information under the Australian Privacy Principles?
What good answers sound like
You’re not looking for jargon. You’re looking for specifics. ‘It reads the job number, customer name and line items from your job system, and nothing else’ is a good answer. ‘Don’t worry, it’s all secure’ is not.
If a provider can’t tell you plainly what their system touches, where the data goes and how to switch it off, wait until they can.
How we approach it
Each automation we build follows your rules, flags anything unusual for a person to check, and only connects to the systems and data it needs. We’re happy to go through every question on this list during the free 15-minute audit call, before anything is built.